Showing results for 
Search instead for 
Did you mean: 

Tunning APIC-EM


We started to use APIC-EM for massive hardware lifecycle and now we have more that 1000 devices in inventory. 

APIC-EM server is a VM : 12 vCPU's, 64 GB RAM and 500 GB HDD. APIC-EM Version  However after about 700 devices are being provisioned system performance started to degrade. Now I see that  CPU is 100 % and RAM almost 100% in use.


Question: is there way to tune APIC-EM? Turn off all services except PNP, device discovery and device inventory? 


Another question might be related to that issue : As number of APIC-EM administrators increased so someone removed Project with 70% of all provisioned devices and now all these devices are in Unclaimed state. Is there way to bring them back via API somehow ? 

Seb Rupik
VIP Advisor

Hi there,

I can't help with your performance question, but I would instinctively say no. @aradford might be able to advise when he's online.


Regarding the the lost project issue. If you do a GET /pnp-search with a 'state': 'UNCLAIMED', do the device objects contain a 'site' object referencing the old project name? If so, try recreating the old project with exactly the same name and it *may* repopulate. 

I had a similar experience with APIC-EM when I was using it a while back, and it seemed very sloppy in the way it tidied objects when parent groups where removed: might be lucky!




Unfortunately this doesn't work. I created same project name but all devices appeared in Unclaimed state.

So I was able to add Unclaimed device  to Project from API using :  POST /pnp-project/{projectId}/device. I was using payload '


[{"serialNumber": "FCW2XXXXXX", "hostName": "switch-1", "platformId": "WS-C2960L-48PS-LL", "state":"PROVISIONED"}]

But state of this devices changed to Error  - "ERROR_DURING_TRUSTPOINT_PROVISIONING"  I'm wonder is there a way to change this status to Provisioned ? 


I tried to update device using  PUT /pnp-project/{projectId}/device using payload : 



But nothing changed. 






Cisco Employee

re performance.  What is the Disk IO?


Most of the issues i have seen have been IO related.


If you run the following command on the APIC-EM CLi.  you should get the disk IO.  Below is a bad example of 23M of IO.


$ dd if=/dev/zero of=/tmp/foo bs=1M count=512 conv=fdatasync

512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 23.7006 s, 22.7 MB/s

Got this:


512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 1.19586 s, 449 MB/s

I think it is not too bad.

Yeap… that looks fine. Just needs to be more than 200M.

I have customers who have loaded > 1000 switches with pnp.

Can you open a TAC case?

I tried, but I don't have support contract for that. 

Content for Community-Ad